ai-create
Pass
Audited by Gen Agent Trust Hub on Nov 9, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill limits tool access to a strict whitelist, specifically permitting only
dateandmkdiroperations via Bash, which prevents arbitrary command execution. - [SAFE]: Filesystem safety is enforced through explicit rules that prevent writing outside the project root without user confirmation.
- [INDIRECT_PROMPT_INJECTION]: The skill reads project context files (e.g., README.md, AGENTS.md) and user arguments to populate the generated folder metadata. This creates a potential surface for indirect prompt injection. Evidence: (1) Ingestion points: $ARGUMENTS and context files identified in SKILL.md Step 1. (2) Boundary markers: absent. (3) Capability inventory: mkdir and file writing in SKILL.md Step 6. (4) Sanitization: absent. The severity is considered safe as these ingestion points are central to the skill's primary purpose and the agent lacks network or high-privilege tools to exploit the injected content.
Audit Metadata